Fig. 3From: Predicting poverty. Data mining approaches to the health and demographic surveillance system in Cuatro Santos, NicaraguaCross-validated conditional inference tree with candidate predictor variables on assets and part of unsatisfied basic need index excluded. Each end node is at least 200 individuals. Black area in end nodes shows proportion of poor (2–4 unsatisfied basic needs) and grey non-poor (0–1 unsatisfied basic needs). All variables measured at household levelBack to article page